Do I need to trim the reads on samples with <0.2% adapter content?

Question

[compuTE]

Hello,

I have a basic question: I ran FastQC and I found that my samples (Illumina TruSeq RNAseq) have less than 0.2% of adapter content. I thought, as this is not even a warning, I could leave it at that and proceed with the mapping of my reads. I'm thinking now that this might have not been the optimal decision. Might be a perfectionist question, but better to ask - should I trim and repeat the mapping?

Thanks,

Answer 1

[liorglic]

I tend to agree with you that when there is very little adapter contamination, trimming them is not a critical step. However, the question is what's the cost of doing that anyway, and not worry about it anymore? The cost is usually disk space and compute time. If you are short on these resources, or have lots and lots of data to process, then maybe skipping the adapter trimming step is not a bad idea. If it won't be too much of a task, then just get it over with. Also keep in mind that adapter trimming is not the only recommended preprocessing step. So if for example you run a tool like Trimmomatic to trim low quality bases/reads, then why not also tell it to trim adapters?

[compuTE]

Thank you for your answer! I agree with you.

Update: I just read that the mapper I used (STAR) does soft-clipping for adapter or poor quality sequences, so it will not be a big difference (if any..). But for anyone looking for an answer of this question, I think the best, as suggested by both answers here, is to just get it over with :-).

Answer 2

[zack.saud]

Depends what your downstream process is. For example, if you are planning to correct long reads with those short reads, then definitely trim them! If you are assembling the short reads only, maybe ask the person who made the assembly tool how tolerant the assembler is to adapter contamination (some assemblers handle them well, others not so well).

Hope this helps

[compuTE]

Thank you! Follow up question: I'm not really going to make transcript assembly with these reads. If I'm only quantifying gene expression, do you think it is going to make a big difference?